Thermodynamics at work: The pressure derivative of the specic heat

J. Guemez
Departamento de Fsica Aplicada, Universidad de Cantabria, E-39005 Santander, Spain

C. Fiolhais and M. Fiolhais

Departamento de Fsica and Centro de Fsica Computacional, Universidade de Coimbra, P-3000 Coimbra, Portugal

Received 25 February 1999; accepted 2 July 1999 Thermodynamics relates measurable quantities such as thermal coefcients and specic heats. The rst law, which implies that the enthalpy is a function of state, yields a relation for the pressure derivative of the specic heat c P . The second law gives a simpler and well-known relation for this pressure derivative. We compare the values of the pressure derivative of c P obtained from the rst and second laws to the values obtained from measurements for water at different pressures. The comparison illustrates the scope and methodology of thermodynamics. 1999 American Association of
Physics Teachers.

I. INTRODUCTION Thermodynamics is based on a small number of principles and provides a formalism which relates the various thermal coefcients and specic heats for any substance in equilibrium.1 Examples of typical thermodynamic relations are the reciprocity theorem, arising from the zeroth law, Reechs relation, arising from the rst law, and Mayers relation, arising from the second law. These relations can be summarized as follows. From the zeroth law, the reciprocity theorem for a PVT pressure, volume, and temperature system states that
TP

and s (1/v )( v / P) s is the adiabatic compressibility.6 Equation 3 implies that the ratio of the specic heats at constant pressure and at constant volume7,8 equals the ratio of the isothermal and adiabatic compressibilities.9 If these quantities are measured independently, we could verify the rst law. Equation 3 can be used to determine the specic heat, c V , which is difcult to measure because of the difculty of keeping the volume constant. Mayers relation follows from the second and rst laws and can be expressed as:10 cV cP T
2 T

Mayers relation .

reciprocity theorem ,

(1/P)( P/ T) V is the relative pressure coefwhere (1/v )( v / T) P is the cubic thermal expansion cient, coefcient, v is the specic volume, and T (1/v ) ( v / P) T is the isothermal compressibility. Equation 1 follows readily from the existence of a thermal equation of state, because
v P

P T

T v

1.
P

It is interesting to show the validity of Eq. 1 using only experimental data. A pedagogical example using a rubber strip a tension, length, and temperature system is discussed in Ref. 2. If Eq. 2 were not experimentally fullled, the system would not be in equilibrium.3 Equation 1 can be used to obtain a coefcient such as , which is difcult to measure, from and T , which are easier to measure. It is difcult in practice to keep the volume constant when the temperature is being changed by heating. We can derive Reechs relation from the rst law,4 which q P d v , where u is the specic can be expressed as du internal energy and q is the specic heat transfer. Reechs relation5 states that cV cP
s T

Equation 4 is one of the most important relations of thermodynamics.11 The independent measurement of the coefcients appearing in Eq. 4 allows us to verify the rst and second laws of thermodynamics. We may also use Eq. 4 , instead of Eq. 3 to obtain c V . Equations 1 , 3 , and 4 imply that among the thermal coefcients , T , s , and , and the specic heats c P and c V , only three can be considered independent. For example, if we choose T , , and c P , then c V could be obtained from Eq. 4 and and s can be obtained from Eqs. 1 and 3 . It is difcult to verify any of the relations, Eqs. 1 , 3 , or 4 directly, because at least one of the quantities in each relation is difcult to measure. However, we can test experimental data for thermodynamic consistency by considering equations for ( c P / P) T . In this paper, we show how to obtain ( c P / P) T using only the rst law. We will arrive at the relation cP P
v
T

cP cV

where the prime denotes differentiation with respect to temperature at constant pressure. If we substitute Eq. 4 in Eq. 5 , we obtain another relation cP P Tv
T 2

Reechs relation ,

where c V ( q/ T) V ( u/ T) V and c P ( q/ T) P ( h/ T) P are the specic heats at constant volume and pressure, respectively, the specic enthalpy h u P v ,
1100 Am. J. Phys. 67 12 , December 1999

Equation 6 is well-known, but Eq. 5 is not. Note that the latter relation provides an experimental check of the rst law. We emphasize that the rst law can be veried independently of the second. Thus, we may imagine a hypothetical reversible world where the rst law holds but not the second.
1999 American Association of Physics Teachers 1100

Table I. Experimental data for water at P 0 101.3 kPa and temperatures ranging from the ice point to the boiling point Ref. 15 . The rst six columns are direct experimental data and the last three columns are determined indirectly. The specic heat in column 8 was obtained from the relation c V c P S / T and the specic heat in column 9 was calculated by using c V c P T 2 v /k T . T C 0 10 20 30 40 50 60 70 80 90 100
v 103 m3 kg 1

106 K
1

1010
1

1010
1

cP J kg
1

Pa

Pa

Eq. 1 K 1 1.3171 1.8162 4.4485 6.6861 8.5975 10.226 11.605 12.760 13.716 14.491 15.110

c V Eq. 3 J kg 1 K 1 4215.1 4187.5 4154.5 4115.9 4072.6 4025.6 3976.0 3924.7 3872.6 3820.5 3768.7

c V Eq. 4 J kg 1 K 1 4215.1 4187.5 4154.5 4115.9 4072.6 4025.6 3976.0 3924.7 3872.6 3820.5 3768.8

1.000 16 1.000 30 1.001 79 1.004 37 1.007 84 1.012 11 1.017 09 1.022 72 1.029 00 1.035 89 1.043 41

67.89 87.96 206.80 303.23 385.30 457.60 523.07 583.74 641.11 696.24 750.30

5.0885 4.7810 4.5891 4.4770 4.4240 4.4174 4.4496 4.5161 4.6143 4.7430 4.9018

5.0855 4.7758 4.5591 4.4100 4.3119 4.2536 4.2281 4.2307 4.2584 4.3093 4.3819

4217.6 4192.1 4181.8 4178.4 4178.5 4180.6 4184.3 4189.5 4196.3 4205.0 4215.9

In Sec. II we derive these two results, and in Sec. III we compare them using experimental data for liquid water. Some conclusions and suggested problems are given in Sec. IV. II. THERMODYNAMIC RELATIONS The rst law asserts that measurements of the variations of the internal energy in an isolated thermal system can be reduced to a pure mechanical problem.12 This statement assumes the existence of adiabatic walls. Furthermore, these variations are independent of the process from the initial to the nal state, and thus the internal energy is a function of state. In practice, the most convenient independent variables are the intensive quantities T and P. 13 For this reason, the enthalpy should be used instead of internal energy. From the rst law, we nd the following expression relating the derivatives of the enthalpy: h T cV v
v

ds

1 dh T

v dP T

1 T

h P

v dP
T

cP dT. T

11

Applying Schwartzs theorem to the entropy, we obtain 1 T cP P 1 T2 h P

v
T

1 2h T T P

1 T

v T

.
P

12 Using Eq. 10 , we obtain h P

v
T

v T

v 1
P

13

This important result may be called the compatibility condition because it equates two different kinds of quantities: a derivative of the enthalpy on the left-hand side and a function obtained directly from the equation of state on the righthand side. Equation 6 is obtained by taking the temperature derivative of Eq. 13 at constant pressure. III. ANALYSIS OF EXPERIMENTAL DATA In the following we obtain the derivative ( c P / P) T for water using data for the thermal coefcients and specic heats and various thermodynamic relations. We have chosen water because accurate data are available for a wide range of temperatures and pressures. The data in Table I are for liquid water at 101.3 kPa normal atmospheric pressure .15 The data in Table II is for liquid water in the pressure range 50 105 Pa to 200 105 Pa. 16 The specic volume, v , and the thermal expansion coefcient, , given in Table I can be measured directly. To obtain T , specic volumes at different pressures are needed see Table II . The coefcient s is measured from the speed of sound. The enthalpy, h, and the specic heat, c P , are measured using an electric calorimeter. On the other hand, and c V are difcult to measure directly, because the volume of water changes upon heating. It is convenient, therefore, to use, respectively, the reciprocity theorem, Eq. 1 and Eq. 3 to obtain them. Table I shows the resulting values for and c V . In the last column of Table I we show the values of c V as calculated from Eq. 4 . The values of c V turn out to be the same within 5 digits using either relation.
Guemez, Fiolhais, and Fiolhais 1101

P T

h T

h P

P T

.
v

Using the denitions of c P and cP cV

T

and Eq. 1 , we obtain 8

h P

,
T

which may be written as: h P

v
T

cP cV

Taking the derivative of this expression with respect to the temperature and applying Schwartzs theorem the value of mixed derivatives is independent of the order in which the derivatives are taken to the enthalpy leads directly to h T P
2

cP cV

h P T

cP P

,
T

10

which is the same as Eq. 5 . This result was given in a different form by Max Planck in his classical monograph.14 In Appendix A, we derive Plancks result and show its equivalence to Eq. 5 . Although Eq. 6 can be obtained by substituting Eq. 4 into Eq. 5 , another way of deriving Eq. 6 is the following. If we combine the second and the rst laws, we have
1101 Am. J. Phys., Vol. 67, No. 12, December 1999

Table II. Experimental specic volumes, v 103 in m3 kg different pressures and temperatures in C Ref. 16 . P 5 MPa T 0 20 40 60 80 100
v

, and enthalpies, h 10

in J kg

, for water at

P 10 MPa h
v

P 15 MPa
v

P 20 MPa
v

h 10.04 93.33 176.38 259.49 342.84 426.51

h 15.04 97.99 180.78 263.67 346.81 430.28

h 20.00 102.58 185.15 267.85 350.80 434.06

0.9977 0.9995 1.0056 1.0149 1.0268 1.0410

5.03 88.65 171.98 255.30 338.85 422.81

0.9952 0.9972 1.0034 1.0127 1.0245 1.0385

0.9928 0.9950 1.0013 1.0105 1.0222 1.0361

0.9904 0.9928 0.9992 1.0084 1.0199 1.0337

To obtain ( c P / P) T from experimental results at different pressures, we rst consider the enthalpies given in Table II. We used a ve point algorithm for numerical derivatives see Appendix B to compute c P ( h/ T) P for each T and P of Table II. For the range of pressures considered, c P is a linear function of P for each T. To reduce the numerical error in ( c P / P) T , we made linear ts to c P ( P) see Appendix B for each T. The values of c P at atmospheric pressure in Table I are included in these ts. The result is shown in Table III. The values of ( c P / P) T are negative, independent of the pressure. They are displayed in the second column of Table IV, which includes interpolated values for intermediate temperatures. To compute ( c P / P) T at atmospheric pressure using Eq. 5 , the derivative on the right-hand side with respect to T at constant P may be evaluated from the data of Table I using the same ve points algorithm. The numerical derivative on the right-hand side of Eq. 5 , with c V given by Eq. 3 , was taken after the expression in brackets was evaluated. The result appears in the third column of Table IV. Because the experimental data in Table I satisfy Mayers relation for c V , we can also evaluate ( c P / P) T using Eq. numerically from 6 . In this case we have to obtain given in Table I. The result for ( c P / P) T , using the ve point algorithm, is the fourth column of Table IV. We take the latter result as the standard one because less numerical work is required besides the empirical errors, only the nuis present. The maximum relative deviamerical error of tion of the results for Eq. 5 relative to Eq. 6 is 0.6%. However, the maximal deviation between the second and the fourth columns of Table IV is 16% and occurs at T 0 C. These deviations are due to the inaccuracy in the derivatives of h with respect to the temperature the error is bigger at the end points , in the linear ts to c P , and in the linear interpolations of ( c P / P) T for intermediate temperatures. A further reason for relying more on the last column of Table
Table III. Fits to c P ( P) obtained from c P ( h/ T) P using the data given in Table II and the values of c P at atmospheric pressure from Table I. T C 0 20 40 60 80 100 4217.2 4183.4 4177.1 4182.7 4196.6 4218.0 cP J kg
1 1

IV than on the other columns is the fact that the enthalpy and its variations are more difcult to measure than the volume and the expansion coefcient. IV. CONCLUSIONS AND PROBLEMS We have used experimental data for liquid water to obtain ( c P / P) T using the rst and second laws and compared the indirect results to direct measurements done at different pressures. We argued that Eq. 6 arising from the second law requires less information and gives more accurate numerical results. The two thermodynamic predictions for ( c P / P) T agree very well with each other and reasonably well with the values extracted from experiments at different pressures. We expect thermodynamic relations to be universal. Because water shows anomalous behavior,17 testing thermodynamic relations using data on water provides a good test of thermodynamics. Similar data sets for other substances may also be used to check their thermodynamic consistency. The questions studied here may help students better understand the formalism of thermodynamics. Students often become lost in the mathematical formalism and might not fully appreciate the physical content of the various thermodynamic relations. In contrast to engineering students, physics majors do not make much use of thermodynamic tables so that the connection of thermodynamics to experiment is limited. We think that it is educational to use empirical data and simple numerical methods to analyze it. To this end, we propose three problems for students.
Table IV. Comparison of values of ( c P / P) T 106 in J kg 1 K 1 Pa 1 for water at 101.3 kPa. The second column was obtained from the slopes of the equations in Table III. The third column is obtained using Eq. 5 rst law taking numerical derivatives of data in Table I. The last column is obtained from Eq. 6 second law using the same table. In the second column, linear interpolation was used for the temperatures not given in Table III. The most accurate results are in the last column. T C 0 10 20 30 40 50 60 70 80 90 100 From Table III 4.171 3.481 2.791 2.595 2.398 2.281 2.164 2.129 2.093 2.233 2.372 From Eq. 5 4.935 3.809 3.127 2.715 2.462 2.307 2.219 2.185 2.190 2.233 2.306 From Eq. 6 4.942 3.808 3.118 2.713 2.463 2.308 2.220 2.184 2.187 2.228 2.319

4.1714 2.7913 2.3982 2.1642 2.0930 2.3722

10 10 10 10 10 10

6 6 6 6 6 6

P P P P P P

1102

Am. J. Phys., Vol. 67, No. 12, December 1999

Guemez, Fiolhais, and Fiolhais

1102

Table V. Experimental data of v and for water at 101.3 kPa for temperatures close to 0 C to be used in Problem 1 Ref. 18 . T C 0 1 2 3 4 5
v 103 m3 kg 1

1.000 160 1.000 100 1.000 060 1.000 036 1.000 028 1.000 036

By applying Schwartzs theorem to T T( P, v ), we obtain Plancks result 14 . Now we use T and P as independent variables to verify that Eq. 18 is equivalent to Eq. 5 . From the denitions of and T and Eq. 1 , we have cV v Moreover, T v P
2 T T P

T P

cV T

T v

T P

cv
2
v

19

T v

1 v

, 20

1 Water has a density maximum at 4 C Table V . Use the ve points algorithm given in Appendix B to evalufor the temperatures indicated in the table. ate and Evaluate the right-hand side of Eq. 6 at 0 C and compare the result with the corresponding ones in Table IV. 2 From the data of Table II obtain ( h/ P) T by tting values at the same temperature and different pressures. Then evaluate the same quantity at atmospheric pressure from Eq. 9 rst law and from Eq. 13 second law using the data of Table I. 3 Insert the compatibility condition, Eq. 13 , in Eq. 11 . Use a numerical integration algorithm,19 to integrate Eq. 11 and construct a table for the entropy s(T, P) from the data in Table II. Take s(T 273.15 K, P 5.00 MPa) 0.0001 kJ kg 1 K 1. 16

and the left-hand side of Eq. 18 becomes cV

T 2

cV v T v

1 cV v cP P

21

On the other hand, we have cP P cP

T T

1 , v

22

and the right-hand side of Eq. 18 is 1 v cP P cP

T T T

cP v 1 v

1 cP v

cP P

1.
T

23

APPENDIX A: PLANCKS RESULT Based on the rst law only, Planck established the following relation between thermal coefcients and specic heat:14 cP cV 1 T P v
2

If we equate Eqs. 21 and 23 and reorder terms, we obtain Eq. 5 . APPENDIX B: NUMERICAL METHODS Given N data points for the quantity y(x), the derivative dy/dx can be obtained numerically by using the ve point algorithm:20 dy dx dy dx dy dx dy dx 1 24 x 1 24 x 50y 1 96y 2 72y 3 32y 4 6y 5 , 6y 1 20y 2 36y 3 12y 4 2y 5 , 16y i 16y i 2y i ,

cP P

T v

cV v

T P

Plancks result .

14

Here we derive this result and show that it is equivalent to Eq. 5 . Taking the specic internal energy u as a function of the specic volume v and pressure P, we may write u P From du u v cV
v

T P

.
v

15

q cP
P

P d v , we obtain T v P.
P

1 2y i 24 x 1 24 x 20y N

16

2y N
1

12y N

36y N

N 1

Taking derivatives of Eqs. 15 and 16 and applying Schwartzs theorem, we obtain

v

6y N ,
4

cV

T P

cP

T v

dy dx

P
P
v

17

1 6y N 24 x 96y N
1

32y N

72y N

50y N .

or cV v T P cV
v

T v P

cP P

T v

cP
P

T P v

1. 18

The algorithm also applies to partial derivatives. This algorithm was implemented using Excel. We used the analytical equation of state of Thomsen and at T 0 C and P Hartka21 to estimate the error for 101.3 kPa. For a temperature step of 10 C, the result
Guemez, Fiolhais, and Fiolhais 1103

1103

Am. J. Phys., Vol. 67, No. 12, December 1999

(T 0 C) 16.0001 10 6 K 2 obtained with the ve point algorithm is accurate to within 2 10 10 K 2. Excel was also used to t data for c P ( P) to a straight line linear regression . The use of spreadsheets is appropriate for the type of analysis presented in this paper and requires only modest knowledge of computer programming.
M. W. Zemansky and R. H. Dittman, Heat and Thermodynamics McGraw-Hill, New York, 1997 , 7th ed., Chap. 9. This book was reviewed in Am. J. Phys. 66, 164166 1998 . 2 H. B. Carroll, M. Eisner, and R. M. Henson, Rubber band experiment in thermodynamics, Am. J. Phys. 31, 808808 1963 . 3 B. Bergthorsson, Temperature, transitivity, and the zeroth law, Am. J. Phys. 45, 270271 1977 . 4 F. W. Sears, and G. Salinger, Thermodynamics, Kinetic Theory, and Statistical Thermodynamics Addison-Wesley, Reading, MA, 1975 , pp. 101 102. 5 C. Truesdell, The Tragicomical History of Thermodynamics 18221854 Springer, New York, 1980 , p. 189. Ferdinand Reech 18051884 was a French naval engineer, whose name is associated with Eq. 3 in the French literature. 6 The subscript s simply means that the process is adiabatic and reversible; we do not need the notion of entropy here. See Ref. 1, p. 239. 7 K. Weltner, Measurement of specic heat capacity of air, Am. J. Phys. 61, 661662 1993 . 8 D. J. Cronin, The temperature variation of gamma for various gases: a student experiment, Am. J. Phys. 32, 700704 1964 . 9 S. Velasco, F. L. Roman, and J. Faro, A simple experiment for measuring the adiabatic coefcient of air, Am. J. Phys. 66, 642643 1998 . 10 Reference 1, p. 270. 11 G. Weinreich, Fundamental Thermodynamics Addison-Wesley, Reading, MA, 1968 , notes that The fact that such a relation is mathematically an almost trivial identity should not blind us to the fact that there is nothing
1

trivial physically about it or about other thermodynamic identities such as the Maxwell relations. The experiments which are used to obtain heat capacities are completely different from those that give the expansion coefcients. That the resulting numerical values consistently satisfy that relation must, accordingly, be viewed as the working of a deep physical law. 12 A. J. Mallinckrodt and H. S. Leff, All about work, Am. J. Phys. 60, 356365 1992 . 13 Furthermore, for water at normal pressure and in the temperature range 08 C, v and P cannot be independent variables because there are two values of T corresponding to the same v . 14 M. Planck, Treatise on Thermodynamics Dover, Mineola, 1990 , p. 59, who wrote This equation contains only quantities which may be experimentally determined, and therefore furnishes a means for testing the rst law of thermodynamics by observations on any homogeneous substance. 15 Reference 1, p. 242. 16 J. H. Keenan et al., Steam Tables. Thermodynamic Properties of Water Including Vapor, Liquid, and Solid Phases Wiley, New York, 1969 , pp. 104105. These data are transcribed in text books, such as A. Shavit and C. Gutnger, Thermodynamics. From Concepts to Applications Prentice Hall, Europe, 1995 , pp. 452453. However, one must be aware of transcription errors. We found two in the latter book: one for v (T 100 C, P 5.00 MPa) and the other for v (T 20 C, P 200 MPa). 17 H. Tanaka, Simple physical explanation of the unusual thermodynamic behavior of liquid water, Phys. Rev. Lett. 80, 57505753 1998 . 18 E. P. Gyftopoulos and G. P. Beretta, Thermodynamics: Foundations and Applications MacMillan, New York, 1991 , pp. 642645. 19 F. S. Acton, Numerical Methods that Work The Mathematical Association of America, Washington, DC, 1990 , Chap. 10. 20 Handbook of Mathematical Functions with Formulas, Graphs, and Mathematical Tables, edited by M. Abramowitz and I. A. Stegun Dover, New York, 1972 , p. 914. 21 J. S. Thomsen and T. J. Hartka, Strange Carnot cycles; Thermodynamics of a system with a density extremum, Am. J. Phys. 30, 2633 1962 .

IS QUANTUM THEORY ROTTEN? I also told Wheeler that from the time that Bell began to study the quantum theory, he had conceptual problems with it, and that I had asked Bell if, at that time, he thought that the theory might simply be wrongto which Bell had answered, I hesitated to think it might be wrong, but I knew that it was rotten.
Jeremy Bernstein, John Wheeler: Retarded Learner, in Quantum Proles Princeton University Press, Princeton, 1991 , p. 137.

1104

Am. J. Phys., Vol. 67, No. 12, December 1999

Guemez, Fiolhais, and Fiolhais

1104